def clustermachine(matrix, distance_metric, clusters=4):
"""
The clustermachine takes a matrix with word freqs and clusters according to the distance_metric.
Clusters sets the input if algorithm needs a pre-determined number of clusters.
Last two will not be used by all algorithms.
"""
no_of_clusters = range(clusters)
result = []
t = time.time()
## # 1: kmeans
# for x in [2,4,6,8,10]:
# model=sklearn.cluster.KMeans(x,tol=0)
# clustering=model.fit(matrix)
# centroids=clustering.cluster_centers_
# labels=clustering.labels_
# inertia=clustering.inertia_
# kmeans=ct.Clustering(model, clustering.labels_, clustering.cluster_centers_)
# result.append(kmeans)
# print [i.name for i in result][len(result)-1], [i.no_of_clusters for i in result][len(result)-1]
# u=time.time()
# print (u-t)/60
# #
# ## #2: MeanShift, takes forever @ 12600, 42
# model=sklearn.cluster.MeanShift()
# clustering=model.fit(matrix)
# centroids=clustering.cluster_centers_
# labels=clustering.labels_
# meanshift=ct.Clustering(model, clustering.labels_, clustering.cluster_centers_)
# result.append(meanshift)
# u=time.time()
# print [i.name for i in result][len(result)-1]
# print (u-t)/60
#
# 3: Affinity Propagation, breaks @ 12600, 42
# model=sklearn.cluster.AffinityPropagation()
# clustering=model.fit(matrix)
# centroid_index=model.cluster_centers_indices_
# centroids=clustering.cluster_centers_
# labels=clustering.labels_
# aff_matrix=clustering.affinity_matrix_
# its= clustering.n_iter_
# affinity=ct.Clustering(model, clustering.labels_, clustering.cluster_centers_)
# result.append(affinity)
# u=time.time()
# print [i.name for i in result][len(result)-1], [i.no_of_clusters for i in result][len(result)-1]
# print (u-t)/60
# ## #4: Spectral clustering
# model=sklearn.cluster.SpectralClustering()
# clustering=model.fit(matrix)
# labels=clustering.labels_
# aff_matrix=clustering.affinity_matrix_
# spectral= ct.Clustering(model, clustering.labels_)
# result.append(spectral)
# u=time.time()
# print [i.name for i in result][len(result)-1], [i.no_of_clusters for i in result][len(result)-1]
# print (u-t)/60
##watch out --------- centroids are indices!!!!!
# ## # 5: DBCASN, takes forever @ 12600, 42
# for x in [2,4,8,16,32]:#[0.175, 0.2, 0.225, 0.3]:
# model=sklearn.cluster.DBSCAN(eps=x, metric=distance_metric, algorithm='brute')
# clustering=model.fit(matrix)
# core_samples=clustering.core_sample_indices_
# #print "core samples", matrix[clustering.core_sample_indices_]
# components=clustering.components_
# print "components", len(components)
# labels=clustering.labels_
# print labels
# dbscan= ct.Clustering(model, clustering.labels_, matrix[clustering.core_sample_indices_])
# result.append(dbscan)
# u=time.time()
# print [i.name for i in result][len(result)-1], [i.no_of_clusters for i in result][len(result)-1]
# print (u-t)/60
#
# ##GUASSIN DOEs NOT FIT OUR SCHEMA AT THIS POINT
# ## 6: GAUSSIAN MIXTURE.
# for x in [2,4,6,8,12,16,20,24]:
# model=sklearn.mixture.DPGMM(x, n_iter=100, verbose=0)
# print "initial weights", model.weights_
# print "initial components", model.n_components
# print "initial converge", model.converged_
# model.fit(matrix)
# print "trained weights", model.weights_
# print "trained components", model.n_components
# print "trained converge", model.converged_
# print "\n predict", model.predict(matrix)
# print "means", model.means_
# #print "\n predict probs", model.predict_proba(matrix)
# dirichlet= ct.Clustering(model, model.fit_predict(matrix), model.means_)
# u=time.time()
# result.append(dirichlet)
# print (u-t)/60
#
# for x in [2,4,8,16,32]:
# model=sklearn.mixture.GMM(x, n_iter=500, verbose=0)
# print "initial weights", model.weights_
# print "initial components", model.n_components
# print "initial converge", model.converged_
# model.fit(matrix)
# print "trained weights", model.weights_
# print "trained components", model.n_components
# print "trained converge", model.converged_
# print "\n predict", model.predict(matrix)
# print "means", model.means_
# #print "\n predict probs", model.predict_proba(matrix)
# gauss= ct.Clustering(model, model.fit_predict(matrix), model.means_)
# u=time.time()
# result.append(gauss)
# print (u-t)/60
#
#
#These are essentially trees; maybe need a different approach. They are kinda predictive
# ## #7: Agglomerative
for x in [4]:
model = sklearn.cluster.AgglomerativeClustering(
affinity=distance_metric, n_clusters=x, linkage='complete')
clustering = model.fit(matrix)
labels = clustering.labels_
leaves = clustering.n_leaves_
children = clustering.children_
components = clustering.n_components_
ward = ct.Clustering(model, clustering.labels_)
result.append(ward)
u = time.time()
print[i.name for i in result
][len(result) - 1], [i.no_of_clusters
for i in result][len(result) - 1]
print(u - t) / 60
# #
#
# print [i.name for i in result][len(result)-1], [i.no_of_clusters for i in result][len(result)-1]
# print (u-t)/60
# model=sklearn.cluster.AgglomerativeClustering(affinity='cosine', linkage='complete')
# clustering=model.fit(matrix)
# labels=clustering.labels_
# leaves=clustering.n_leaves_
# components=clustering.n_components_
# ward= ct.Clustering(model, clustering.labels_)
# result.append(ward)
# u=time.time()
# print [i.name for i in result][len(result)-1], [i.no_of_clusters for i in result][len(result)-1]
# print (u-t)/60
# ## #8: Birch Hierarchical
# model=sklearn.cluster.Birch(threshold=0.025)
# clustering=model.fit(matrix)
# labels=clustering.labels_
# root=clustering.root_
# subcluster_labels=clustering.subcluster_labels_
# birch= ct.Clustering(model, clustering.labels_)
# result.append(birch)
# u=time.time()
# print [i.name for i in result][len(result)-1], [i.no_of_clusters for i in result][len(result)-1]
# print (u-t)/60
return (result)
def clustermachine(matrix, algorithm, clusters=3): #we need a similarity matrix similarity_matrix=metrics.pairwise.euclidean_distances(matrix) #meanshift and kmeans take features #others need distance matrixes no_of_clusters=range(clusters) result=[] ## # 1: kmeans model=sklearn.cluster.KMeans(clusters) clustering=model.fit(matrix) centroids=clustering.cluster_centers_ labels=clustering.labels_ inertia=clustering.inertia_ kmeans=ct.Clustering(matrix, model, clustering.labels_, clustering.cluster_centers_) result.append(kmeans) ## #2: MeanShift # model=sklearn.cluster.MeanShift() # clustering=model.fit(matrix) # centroids=clustering.cluster_centers_ # labels=clustering.labels_ # meanshift=ct.Clustering(matrix, model, clustering.labels_, clustering.cluster_centers_) # result.append(meanshift) # # ## #3: Affinity Propagation # model=sklearn.cluster.AffinityPropagation() # clustering=model.fit(similarity_matrix) # centroid_index=model.cluster_centers_indices_ # centroids=clustering.cluster_centers_ # labels=clustering.labels_ # aff_matrix=clustering.affinity_matrix_ # its= clustering.n_iter_ # affinity=ct.Cluster(matrix, model, clustering.labels_, clustering.cluster_centers_) # result.append(affinity) # ## #4: Spectral clustering # model=sklearn.cluster.SpectralClustering() # clustering=model.fit(similarity_matrix) # labels=clustering.labels_ # aff_matrix=clustering.affinity_matrix_ # spectral= ct.Clustering(matrix, model, clustering.labels_) # result.append(spectral) # # ##watch out --------- centroids are indices!!!!! # ## # 5: DBCASN # model=sklearn.cluster.DBSCAN() # clustering=model.fit(matrix) # core_samples=clustering.core_sample_indices_ # components=clustering.components_ # labels=clustering.labels_ # dbscan= ct.Clustering(matrix, model, clustering.labels_, clustering.core_sample_indices_) # result.append(dbscan) # ##GUASSIN DOEs NOT FIT OUR SCHEMA AT THIS POINT ## 6: GAUSSIAN MIXTURE. eh this does not really fit in here model=sklearn.mixture.GMM() clustering=model.fit(matrix) weights=model.weights_ means=model.means_ covars=model.covars_ converged=clustering.converged_ #These are essentially trees; maybe need a different approach. #They are kinda predictive ## #7: Agglomerative // Ward Hierarchical model=sklearn.cluster.AgglomerativeClustering() # clustering=model.fit(matrix) # labels=clustering.labels_ # leaves=clustering.n_leaves_ # components=clustering.n_components_ # ward= ct.Clustering(matrix, model, clustering.labels_) # result.append(ward) # # ## #8: Birch Hierarchical # model=sklearn.cluster.Birch(threshold=0.025) # clustering=model.fit(matrix) # labels=clustering.labels_ # root=clustering.root_ # subcluster_labels=clustering.subcluster_labels_ # birch= ct.Clustering(matrix, model, clustering.labels_) # result.append(birch) return(result)
def clustermachine(matrix, clusters=4):
#we need a similarity matrix to feed into some of the algos
similarity_matrix = metrics.pairwise.euclidean_distances(matrix)
#meanshift and kmeans take features, others need distance matrixes
no_of_clusters = range(clusters)
result = []
t = time.time()
## # 1: kmeans
model = sklearn.cluster.KMeans(clusters, tol=0)
clustering = model.fit(matrix)
centroids = clustering.cluster_centers_
labels = clustering.labels_
inertia = clustering.inertia_
kmeans = ct.Clustering(model, clustering.labels_,
clustering.cluster_centers_)
result.append(kmeans)
print[i.name for i in result][len(result) - 1]
u = time.time()
print(u - t) / 60
#
###CREATING CLUSTERS
#
#this makes clusters; takes the dataset (matrix) and the algorithm
## # 1: kmeans
model = sklearn.cluster.KMeans(clusters)
clustering = model.fit(matrix)
centroids = clustering.cluster_centers_
labels = clustering.labels_
inertia = clustering.inertia_
kmeans = ct.Clustering(model, clustering.labels_,
clustering.cluster_centers_)
result.append(kmeans)
print[i.name for i in result][len(result) - 1]
u = time.time()
print(u - t) / 60
model = sklearn.cluster.KMeans(8)
clustering = model.fit(matrix)
centroids = clustering.cluster_centers_
labels = clustering.labels_
inertia = clustering.inertia_
kmeans2 = ct.Clustering(model, clustering.labels_,
clustering.cluster_centers_)
result.append(kmeans2)
print[i.name for i in result][len(result) - 1]
u = time.time()
print(u - t) / 60
# ## #2: MeanShift, takes forever @ 12600, 42
# model=sklearn.cluster.MeanShift()
# clustering=model.fit(matrix)
# centroids=clustering.cluster_centers_
# labels=clustering.labels_
# meanshift=ct.Clustering(model, clustering.labels_, clustering.cluster_centers_)
# result.append(meanshift)
# u=time.time()
# print [i.name for i in result][len(result)-1]
# print (u-t)/60
#
# 3: Affinity Propagation, breaks @ 12600, 42
model = sklearn.cluster.AffinityPropagation()
clustering = model.fit(matrix)
centroid_index = model.cluster_centers_indices_
centroids = clustering.cluster_centers_
labels = clustering.labels_
aff_matrix = clustering.affinity_matrix_
its = clustering.n_iter_
affinity = ct.Clustering(model, clustering.labels_,
clustering.cluster_centers_)
result.append(affinity)
u = time.time()
print[i.name for i in result][len(result) - 1]
print(u - t) / 60
# ## #4: Spectral clustering
# model=sklearn.cluster.SpectralClustering()
# clustering=model.fit(matrix)
# labels=clustering.labels_
# aff_matrix=clustering.affinity_matrix_
# spectral= ct.Clustering(model, clustering.labels_)
# result.append(spectral)
# u=time.time()
# print [i.name for i in result][len(result)-1]
# print (u-t)/60
##watch out --------- centroids are indices!!!!!
## ## # 5: DBCASN, eanShift, takes forever @ 12600, 42
# model=sklearn.cluster.DBSCAN()
# clustering=model.fit(matrix)
# core_samples=clustering.core_sample_indices_
# components=clustering.components_
# labels=clustering.labels_
# dbscan= ct.Clustering(model, clustering.labels_, clustering.core_sample_indices_)
# result.append(dbscan)
# u=time.time()
# print [i.name for i in result][len(result)-1]
# print (u-t)/60
#
# ##GUASSIN DOEs NOT FIT OUR SCHEMA AT THIS POINT
# ## 6: GAUSSIAN MIXTURE. eh this does not really fit in here
# model=sklearn.mixture.GMM()
# clustering=model.fit(matrix)
# weights=model.weights_
# means=model.means_
# covars=model.covars_
# converged=clustering.converged_
# u=time.time()
# result.append(dbscan)
# print [i.name for i in result][len(result)-1]
# print (u-t)/60
#
#
#These are essentially trees; maybe need a different approach. They are kinda predictive
# ## #7: Agglomerative // Ward Hierarchical
# model=sklearn.cluster.AgglomerativeClustering()
# clustering=model.fit(matrix)
# labels=clustering.labels_
# leaves=clustering.n_leaves_
# components=clustering.n_components_
# ward= ct.Clustering(model, clustering.labels_)
# result.append(ward)
# u=time.time()
# print [i.name for i in result][len(result)-1]
# print (u-t)/60
#
# ## #8: Birch Hierarchical
# model=sklearn.cluster.Birch(threshold=0.025)
# clustering=model.fit(matrix)
# labels=clustering.labels_
# root=clustering.root_
# subcluster_labels=clustering.subcluster_labels_
# birch= ct.Clustering(model, clustering.labels_)
# result.append(birch)
# u=time.time()
# print [i.name for i in result][len(result)-1]
# print (u-t)/60
return (result)
